hustvl
how to use mini dataset
when i run : python tools/data_converter/vad_nuscenes_converter.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag vad_nuscenes --version v1.0-mini --canbus ./data run scueess when run: CUDA_VISIBLE_DEVICES=0 python tools/test.py projects/configs/VAD/VAD_tiny_e2e.py /home/wxl/VAD/ckpts/VAD_tiny.pth --launcher none --eval bbox --tmpdir tmp i has error: No such file or directory: 'data/nuscenes/maps/expansion/boston-seaport.json'
how to use mini dataset
This is what I found to be working for me:
python tools/data_converter/vad_nuscenes_converter.py nuscenes --root-path ../data/nuscenes --out-dir ../data/nuscenes --extra-tag vad_mini --version v1.0-mini --canbus ../data/nuscenes
This will create .pkl files in the nuscenes directory. extra-tag is just a prefix your files will have.
In the config file, e.g. VAD_tiny_e2e.py you have to replace data_root with your path to nuscenes and also change the ann_file (that you created above) in the data dict, e.g. ann_file=data_root + 'vad_mini_infos_temporal_val.pkl'. What I also did in commented out the map_ann_file in the train, val, and test.
You also have to have CanBus extension and a MapExtension for this case.
Then you are free to do the training with:
python tools/train.py projects/configs/VAD/VAD_tiny_e2e.py --launcher none --work-dir ../tmp/VAD
How do I split trainval and test if I want to use a mini data set
I was wondering if you've found a solution to this issue
